In a conventional electronic clinical thermometer for measuring the temperature of a living body, the thermometer is programmed to incorporate a prediction formula set up to perfectly define temperature rise curves, and a so-called "add-on value" determined by the prediction formula is added to an actually sensed temperature to obtain an early display of what the equilibrium temperature should eventually be. To this end, it is required that constants (parameters) used in the prediction formula be set to values which will statistically minimize a prediction error. This is done in the manufacturing process of each electronic clinical thermometer by applying statistical processing to actual measurement values obtained from a temperature probe used in actually measuring temperature.
It is known that temperature rise curves differ from one individual to another, and that a temperature rise curve when temperature is sensed in an armpit will differ considerably from that when temperature is sensed orally even for one and the same individual. As a result, an early display of an accurately predicted equilibrium temperature cannot be obtained in actual practice even if the prediction formula is corrected for dispersion exhibited by the thermal characteristics of the probe.
An electronic clinical thermometer disclosed in the specification of Japanese Patent Application Laid-Open (KOKAI) No. 58-225336 solves the above problem by incorporating a plurality of prediction formulae. More specifically, the thermometer is provided beforehand with a plurality of prediction formulae stipulated by statistical processing based on a large quantity of measurement results. When temperature is actually measured, condition settings are altered on a trial-and-error basis, such as by comparing the rise curve of the temperature being measured and a selected one of the prediction formulae. In other words, the parameters in the selected prediction formula are modified by trial and error to solve the aforementioned problem encountered in the prior art. However, since the plural prediction formulae must be defined in advance, a practical problem that cannot be avoided is the trouble involved in adjusting for dispersion in the thermal characteristics of the temperature probes when these probes are mass-produced. Furthermore, in order to raise the accuracy of prediction, a large number of prediction formulae having different rise curves must be incorporated in the thermometer in advance. Moreover, if an improper prediction formula is selected just as temperature is starting to rise, the transition exhibited by the predicted value may overshoot.
An electronic clinical thermometer disclosed in the specification of Japanese Patent Application Laid-Open (KOKAI) No. 59-187233 solves this problem by setting up a prediction formula which conforms to the rise curve of the actually measured temperature. In other words, use is made of the fact that a linear relationship (TL=A-.tau.'t) exists between a logarithmic value TL of the time differential of measured body temperature and a sampling time t, with A and .tau.' being determined by a regression method. However, since the logarithmic value TL is not measured body temperature per se, an error due to differential and logarithmic calculations is introduced into the temperature data, and the error has a major influence on the setting of the constants A and .tau.'. Moreover, if the measured temperature data include a noise component, the latter affects the predictive results in the manner of an exponential function, causing the predicted values to exhibit a highly unstable transition.